* the accumulated blur weight now takes into account how far verts are from the brush, giving more even results
* verts where the weight wasnt found were being ignored, now treat them as zero weight verts.
also use math functions for calc_vp_strength(), and project the vertices as floats rather then ints to get better accuracy, otherwise no functional changes.
* this loop is called multiple times per vertex (not addressed in this commit)
* functions like brush_use_size_pressure(), brush_use_size_pressure() called unified_settings() twice when they didnt need to.
looks like this code cant work right with multiple scenes, added a comment on this - but at least avoid calling unified_settings() multiple times in single functions.
It is just a limitation of multires baker which doesn't deal correct with
baking to subdivision level 0. It was supposed to work with levels on which
sculpt data is affecting on mesh, so interpolation between grids works correct.
Fully accurate baking in this case will need raycasting stuff which will make
it much slower and will remove main benefit of regular baker -- speed and
low memory usage.
Another option would be to make multires apply sculpting data on level 0,
but it's not related at baking at all and has got it's own difficulties.
Notes:
- only cuts through when the occlude geometry button is unpushed,
else the old behavior
- doesn't add points in middle of occluded faces yet, so will not
make partial cuts or hole cuts in occluded faces; for best
results, start and finish cuts completely outside the mesh
- also made slight improvement to behavior when starting cut
off the model
- a small refactor: added knife_add_to_vert_edges for code
used a number of times
* vertex map was a dynamicly realloc'd array when the final size was known, use a fixed array instead.
* vertex map was being calculated even when not used.
* face tesselation was being called twice.
* an unused deform group array was being created.
byte => float, float => float, byte => byte conversions with profile, dither
and predivide. Previously code for this was spread out too much.
There should be no functional changes, this is so the predivide/table/dither
patches can work correctly.
